- Introduction

Always in the context of DataKind’s Fall2024 DataKit Challenge #1 and moving forward from the get started analysis, we will here proceed with the take it further part, investigating the renting availability and its impact on the income, for Orange County in Florida, with the reasoning that renting it’s a faster solution in case of displacements.

#we load the libraries
library(readr)
library(dplyr)
library(ggplot2)
library(DT)
library(tidyr)
library(stringr)
#we load the data frame
data_1_FL_takeitfurther_part1 <- read_csv("data_1_FL_takeitfurther_part1.csv")

#we select the county in exam
county <- "Orange County"

- Assessment of the overall housing stock

We start by enumerating the availability per census tract, as in number of housing units (defined as “anything from a house to an apartment or even a boat if a person is currently living there”).

data_1_FL_takeitfurther_part1 %>%
#we pick the county of choice
  filter(County == county) %>%
#we build the graph, ordering by the Housing Units count  
  ggplot(aes(reorder(`Census Tract`, desc(`Housing Units Count`)), `Housing Units Count`)) +
  geom_col(fill = "#0072B2") + 
  theme(axis.text.x = element_blank(),
        axis.ticks = element_blank()) +
  labs(x = NULL,
       y = NULL, 
       title = "Number of Housing Units per Census Tract for Orange County, Florida")

The graph is ordered by the census tracts with more housing units, to better convey the county average.

We provide a dynamic table (downloadable as an excel file or a pdf) as well to satisfy the need to pinpoint specific instances:

data_1_FL_takeitfurther_part1 %>%
#we pick the county of choice
  filter(County == county) %>%
#we select the variables of interest
  select(1, 3) %>%
#we display the counties with most housing units on top
  arrange(desc(`Housing Units Count`)) %>%
#we transform the Census Tract variable to factor, to more easily select its values
  mutate(`Census Tract` = factor(`Census Tract`)) %>%
#we construct the dynamic table
  datatable(filter = "top",
            rownames = FALSE,
            extensions = "Buttons",
            options = list(dom = 'tpB',
                           buttons = c('excel', 'pdf')))

Out of the overall housing units present, not all are occupied, as we show in the following graph:

#we define a colorblind friendly palette
cbbPalette <- c("#0072B2", "#D55E00")
data_1_FL_takeitfurther_part1 %>%
#we pick the county of choice
  filter(County == county) %>%
#we transform the data into long format, for easier plotting 
  pivot_longer(cols = c(4, 7), names_to = "Status", values_to = "Number of Housing Units") %>%
#we build the graph, ordering by the number of Housing Units
  ggplot(aes(reorder(`Census Tract`, desc(`Number of Housing Units`)), `Number of Housing Units`, fill = Status)) +
  geom_col() +
  scale_fill_manual(values = cbbPalette) +
  theme(legend.position = "bottom",
        legend.title = element_blank(),
        axis.text.x = element_blank(),
        axis.ticks = element_blank()) +
  labs(x = NULL,
       y = NULL,
       title = "Number of Housing Units per Census Tract for Orange County, Florida, \n differentiating by Occupancy")

We can see that there are some spikes in certain census tracts, that we can individuate in the following dynamic table:

data_1_FL_takeitfurther_part1 %>%
#we pick the county of choice
  filter(County == county) %>%
#we calculate the percentage of vacant housing units and we transform census tract into a factor to more easily select its values
  mutate("Vacant %" = round(Vacant / `Housing Units Count` * 100, 2),
         `Census Tract` = factor(`Census Tract`)) %>%
#we select and reorder the variables of interest
  select(1, 3, 4, 7, `Vacant %`) %>%
#we construct the dynamic table  
  datatable(filter = "top", 
            rownames = FALSE, 
            extensions = "Buttons",
            options = list(dom = 'tpB',
                           buttons = c('excel', 'pdf')))

About the occupied housing units, we can differentiate between owned vs rented, and we can see that the split is generally even along the county,

#we define a colorblind friendly palette
cbbPalette <- c("#0072B2", "#56B4E9")
data_1_FL_takeitfurther_part1 %>%
#we pick the county of choice
  filter(County == county) %>%
#we calculate the percentage of occupied housing units that are rented
  mutate(RentProp = `Renter occupied` / Occupied) %>%
#we transform the data into long format, for easier plotting 
  pivot_longer(cols = 5:6, names_to = "Status", values_to = "Number of Housing Units") %>%
#we build the graph, ordering by RentProp
  ggplot(aes(reorder(`Census Tract`, desc(RentProp)), `Number of Housing Units`, fill = Status)) +
  geom_col(position = position_fill()) +
  scale_fill_manual(values = cbbPalette) +
  theme(legend.position = "bottom",
        legend.title = element_blank(),
        axis.text.x = element_blank(),
        axis.ticks = element_blank()) +
  labs(x = NULL,
       y = NULL,
       title = "Proportions of Owner and Renter Occupied Housing Units per Census Tract, \n for Orange County, Florida")

with a slight overall prevalence of ownership.

data_1_FL_takeitfurther_part1 %>%
#we pick the county of choice
  filter(County == county) %>%
#we calculate the total figures
  summarise("Total Occupied" = sum(Occupied),
            "Total Owner" = sum(`Owner occupied`),
            "Total Renter" = sum(`Renter occupied`)) %>%
#we calculate the percentage of owners
  mutate(`Owner %` = round(`Total Owner` / `Total Occupied` * 100, 2))

Some census tracts are very homogeneous, with all renters for example.

data_1_FL_takeitfurther_part1 %>%
#we pick the county of choice
  filter(County == county) %>%
#we calculate the percentage of rented housing units occupied and we transform census tract into a factor to more easily select its values
  mutate(`Renter %` = round(`Renter occupied` / Occupied * 100, 2),
         `Census Tract` = factor(`Census Tract`)) %>%
#we select the variables of interest
  select(1, 4, 5, 6, `Renter %`) %>%
#we display the census tracts with all housing unites occupied by renters on top
  arrange(desc(`Renter %`)) %>%
#we construct the dynamic table
  datatable(filter = "top",
            rownames = FALSE,
            extensions = "Buttons",
            options = list(dom = 'tpB',
                           buttons = c('excel', 'pdf')))

- Renting availability and impact on the income.

Among the vacant housing units, the ones available to rent represent a large number in several census tracts:

#we define a colorblind friendly palette
cbbPalette <- c("#F0E442", "#D55E00")
data_1_FL_takeitfurther_part1 %>%
#we pick the county of choice
  filter(County == county) %>%
#we calculate the percentage of vacant housing units that are available to rent and we collapse the other instances
  mutate(ToRentProp = `For rent` / Vacant,
         Other = rowSums(pick(9:14))) %>%
#we transform the data into long format, for easier plotting
  pivot_longer(cols = c(8, last_col()), names_to = "Status", values_to = "Number of Housing Units") %>%
#we build the plot, ordering by ToRentProp
  ggplot(aes(reorder(`Census Tract`, desc(ToRentProp)), `Number of Housing Units`, fill = Status)) +
  geom_col(position = position_fill()) +
  scale_fill_manual(values = cbbPalette) +
  theme(legend.position = "bottom",
        legend.title = element_blank(),
        axis.text.x = element_blank(),
        axis.ticks = element_blank()) +
  labs(x = NULL,
       y = NULL,
       title = "Vacancy Status per Census Tract for Orange County, Florida")

But there are also some instances where they are absent:

data_1_FL_takeitfurther_part1 %>%
#we pick the county of choice
  filter(County == county) %>%
#we calculate the percentage of vacant housing units that are available to rent and we transform census tract into a factor to more easily select its values
  mutate(`For Rent %` = round(`For rent` / Vacant * 100, 2),
         `Census Tract` = factor(`Census Tract`)) %>%
#we select the variables of interest
  select(1, 7:14, last_col()) %>%
#we display the census tracts with no vacant housing units available to rent on top
  arrange(`For Rent %`) %>%
#we construct the dynamic table
  datatable(filter = "top",
            rownames = FALSE,
            extensions = "Buttons",
            options = list(dom = 'tpB',
                           buttons = c('excel', 'pdf')))

Lastly, we can study how the rents respond to the economic possibilities of the population, starting with the contract rents, defined as monthly rent agreed to without adjustments for utilities or other payments.

data_1_FL_takeitfurther_part1 %>%
#we pick the county of choice
  filter(County == county) %>%
#we transform the data into long format, for easier plotting
  pivot_longer(cols = contains("contract"), names_to = "Rent", values_to = "Number") %>%
#we transform the Rent variable for an easier visualization
  mutate(Rent = str_sub(Rent, 1, -11),
         Rent = factor(Rent, levels = unique(Rent), ordered = TRUE)) %>%
#we build the graph  
  ggplot(aes(Rent, Number)) +
  geom_col(fill = "#0072B2") +
  coord_flip() +
  labs(x = NULL,
       y = NULL,
       title = "Distribution of Contract Rent for Orange County, Florida")

From the graph we can see that the vast majority of contract rents are above $1000.
As a quick comparison, we present the distribution of the average monthly income per household (calculated from the yearly median one).

#we load the income data frame
data_1_FL_getstarted <- read_csv("data_1_FL_getstarted.csv")
data_1_FL_getstarted %>%
#we pick the county of choice
  filter(county == county) %>%
#we change the negative values (present with missing data) to NAs  
  mutate(across(everything(), ~ case_when(.x >=0 ~ .x))) %>%
#we build the graph  
  ggplot(aes(med_hh_inc_est / 12)) +
  geom_density(fill = "#0072B2") +
  scale_x_continuous(labels = scales::label_dollar()) +
  scale_y_continuous(labels = NULL, breaks = NULL) +
  labs(x = NULL,
       y = NULL,
       title ="Distribution of Average Monthly Income per Census Tract for Orange County, Florida")

As usual, we provide as well a dynamic table to complete the graph.

data_1_FL_takeitfurther_part1 %>%
#we pick the county of choice
  filter(County == county) %>%
#we select the variables of interest
  select(1, contains("contract")) %>%
#we transform the Census Tract variable into a factor to easily select between its values
  mutate(`Census Tract` = factor(`Census Tract`)) %>%
#we construct the dynamic table
  datatable(filter = "top",
            rownames = FALSE,
            extensions = c("Buttons", "FixedColumns"),
            options = list(dom = 'tpB',
                           buttons = c('excel', 'pdf'),
                           scrollX = TRUE,
                           fixedColumns = list(leftColumns = 1)))

We can perform the same analysis for gross rent, defined as contract rent plus the average cost of the utilities (electricity, gas, and water and sewer) and fuel (oil, coal, kerosene, wood, etc), noting an obvious shift to more expensive rents.

data_1_FL_takeitfurther_part1 %>%
#we pick the county of choice
  filter(County == county) %>%
#we transform the data into long format, for easier plotting   
  pivot_longer(cols = contains("gross"), names_to = "Rent", values_to = "Number") %>%
#we transform the Rent variable for an easier visualization
  mutate(Rent = str_sub(Rent, 1, -8),
         Rent = factor(Rent, levels = unique(Rent), ordered = TRUE)) %>%
#we build the graph    
  ggplot(aes(Rent, Number)) +
  geom_col(fill = "#0072B2") +
  coord_flip() +
  labs(x = NULL,
       y = NULL,
       title = "Distribution of Gross Rent for Orange County, Florida")

data_1_FL_takeitfurther_part1 %>%
#we pick the county of choice
  filter(County == county) %>%
#we select the variable of interest
  select(1, contains("gross")) %>%
#we transform the Census Tract variable into a factor to easily select between its values
  mutate(`Census Tract` = factor(`Census Tract`)) %>%
#we construct the dynamic table
  datatable(filter = "top",
            rownames = FALSE,
            extensions = c("Buttons", "FixedColumns"),
            options = list(dom = 'tpB',
                           buttons = c('excel', 'pdf'),
                           scrollX = TRUE,
                           fixedColumns = list(leftColumns = 1)))

Census provides us also with the percentage of income spent on gross rent, which surely paints a worse picture of what we could gather from our previous graphs.

data_1_FL_takeitfurther_part1 %>%
#we pick the county of choice
  filter(County == county) %>%
#we transform the data into long format, for easier plotting  
  pivot_longer(cols = 65:last_col(), names_to = "RentPerc", values_to = "Number") %>%
#we transform the Census Tract variable into a factor to better visualize it on the graph
  mutate(Rent = factor(RentPerc, levels = unique(RentPerc), ordered = TRUE)) %>%
#we build the graph  
  ggplot(aes(RentPerc, Number)) +
  geom_col(fill = "#0072B2") +
  coord_flip() +
  labs(x = NULL,
       y = NULL,
       title = "Distribution of Percentage of Income spent on Gross Rent \n for Orange County, Florida")

data_1_FL_takeitfurther_part1 %>%
#we pick the county of choice
  filter(County == county) %>%
#we select the variables of interest
  select(1, 65:last_col()) %>%
#we transform the Census Tract variable into a factor to easily select between its values
  mutate(`Census Tract` = factor(`Census Tract`)) %>%
#we construct the dynamic table
  datatable(filter = "top",
            rownames = FALSE,
            extensions = c("Buttons", "FixedColumns"),
            options = list(dom = 'tpB',
                           buttons = c('excel', 'pdf'),
                           scrollX = TRUE,
                           fixedColumns = list(leftColumns = 1)))